Murtilla fruits (Ugni molinae Turcz.) were obtained from wild and cultivated plants. The wild fruits were obtained from native vegetation in Puerto Saavedra (38°, 45’ S, 73° 21’ W), La Araucanía Region, Chile. The cultivated fruits (14-4 genotype) were grown in the germ plasm bank at the experimental station of Instituto de Investigaciones Agropecuarias (INIA) at Puerto Saavedra. The wild and cultivated fruits were harvested on April 18, 2011 and dried on April 19, 2011 using an industrial drying cabinet (1.8 m long, 2.2 m high, 3.5 m deep) at a constant temperature of 70 °C for 6 hours until a final moisture content of 5.4% was reached, according to Alfaro et al. (2014). The fruits were vacuum packed and protected from light and oxygen. The hydro-alcoholic extracts were prepared according to the method described by Ribeiro et al. (2008) with certain modifications. Different concentrations of ethanol:water and different temperatures were used throughout the experiments (see Table 1 for the conditions) in an attempt to improve the extraction of polyphenols. The dehydrated fruit was ground, and 1.0 g of the samples was placed in a 250 mL Erlenmeyer with 30 mL of solvent (v/v, ethanol/water) and heated to a specific temperature (see Table 1), which was maintained for 50 minutes under agitation in a water bath (Model NI 1322-220, Nova Instrumentos, Brazil). The extracts were then centrifuged at 2057 g for 15 minutes, filtered using Whatman Nº 2 filter paper, and stored under refrigeration at 7 °C in amber flasks until analysis.

The Response Surface Methodology and a multiple regression analysis were performed to determine the effects of the solvent ratio (ethanol:water) and temperature on the total phenolic content (dependent variable) in the murtilla fruit extracts to identify the best conditions for obtaining high antioxidant activity. Adjustments were performed to generate a second order mathematical model that included linear and quadratic terms as well as their interactions. To estimate the experimental error, four assays with coded values related to zero levels, which represented the central point, were performed. To complete the design, trials referencing the axial points positioned at a distance α (α = 2^n/4) from the central point were conducted. At the end of the design, twelve trials had been performed. The experimental design is shown in Table 1.

The mathematical model was evaluated based on the coefficient of determination (R²) and F test, which were determined by a variance analysis using Statistica 11 software (Stat Soft, 2013). The general equation is Y = b₀ + b₁x₁ + b₂x₂ + b₁₁x₁ ² + b₂ ²x₂ ² + b₁₂x₁x₂, where Y is the dependent variable, x₁ and x₂ are the coded levels of the explanatory variables, b₀ is the central point and b’_s is the estimated coefficient from the minimum square method. The total phenolic content of the extracts was the dependent variable in the experimental design. The measurements were determined in triplicate according to the method described by Singleton et al. (1999) using the Folin-Ciocalteu solution as a reagent and gallic acid as a standard. The hydroalcoholic extracts were diluted in ethanol at concentration of 1:10 (extract: ethanol, v/v). An aliquot of 0.5 mL of the diluted sample was transferred to a test tube, and 2.5 mL of the Folin-Ciocalteu reagent:water solution (1:10, v/v) were added. The mixture was vortexed at room temperature in the dark, and after a 5 minute resting period, 2.0 mL of a sodium carbonate 4% (m/v) solution was added, and the mixture was agitated again and kept at rest for two hours at room temperature and protected from light. The absorbance was read at 740 nm using a UV-1203 spectrophotometer (Shimadzu Corporation, Japan). The results were calculated using the standard curve of gallic acid with known concentrations (2.5 to 50 μg·mL^–1), and the results were expressed in mg of gallic acid equivalent (GAE) of dry matter (g).

Ten treatments were established to evaluate the antioxidant ability to prevent lipid oxidation via accelerated stability tests. These treatments included the addition of 50, 100, 150 and 200 mg GAE kg·oil⁻¹ of the wild and 14-4 genotype extracts of murtilla to soybean oil and 200 mg of TBHQ kg·oil⁻¹ to soybean oil. The control treatment consisted of refined soybean oil alone. The total phenolic contents (determined by the Folin-Ciocalteu method) were used to calculate the concentrations of the added extracts (μg GAE·mL^–1).

For each trial, the samples were prepared in triplicate, binned in 50 mL beakers and randomly distributed in the oven. The established temperature was 63 ± 2 ºC and after 36, 72 and 96 hours, the samples were removed for peroxide and ultraviolet absorption analyses.

The peroxide value was determined according to the Cd 8b-90 method (American Oil Chemists’ Society, 2009), and the results are expressed in meq O₂ ·kg^-1 oil. The specific absorbance was determined according the Ch 5–91 method (American Oil Chemists’ Society, 2009), and the results were expressed in E^1% _{1 cm}. The absorbance was read at 232 nm using a UV-1203 spectrophotometer (Shimadzu Corporation, Japan).

The discriminative test was conducted using an unstructured scale with 6 trained tasters to evaluate the odor (rancidity). For this test, 20 mL from each treatment were kept in black plastic cups labeled with three-digit numbers and randomly distributed in the oven at 43 ºC (Institute of Food Technologists, 1981). The complete block design was applied, and the results were analyzed using the statistics discussed in section 2.6.

All evaluations were conducted in triplicate, and the results were expressed as the mean ± standard deviation. The data were statistically analyzed via an analysis of variance (ANOVA) and Tukey’s test (α = 95%) using SAS (Statistical Analysis System, USA) software.

The total phenolic contents of the murtilla fruit extracts using different extraction conditions are shown in Table 1. The estimated effects on the total phenolic content for each murtilla fruit extract (wild and 14-4 genotype) are shown in Table 2.

According to Table 2, the extraction process for both samples of murtilla were affected by the solvent because the linear and quadratic effects were statistically significant. However, the effects of increases in temperature, which promoted an increase in the total phenolic content of the extract were not statistically significant (p < 0.05).

Despite the differences in the results for each trial, both murtilla fruit extracts (wild and 14-4 genotype) showed the same relationship between the total phenolic content and the solvent. After the effects analysis, terms that were not significant were eliminated, and second-order regression coefficients were obtained.

The second-order polynomial equation for the total phenolic content (TPC) for wild murtilla is:

and the mathematical model for the 14-4 genotype murtilla is:

The analysis of variance (Tables 3 and 4) demonstrated that the proposed models were valid because the regressions were statistically significant. The F test showed that the lack of fit was not statistically significant (p < 0.05). The coefficients of determination (R²) were higher than 0.90. Thus, the ANOVA results indicated that the F values calculated from the regressions of both models were three times higher than the F values that were tabulated. Therefore, the mathematical models were valid, and they can be used for predictive purposes.

The adjusted response surfaces of the mathematical models for murtilla fruit extracts from wild and 14-4 genotypes are shown in Figure 1. An ethanol concentration of 49.5% (v/v, ethanol:water) under all temperatures provided a higher yield of total phenolic content (18.39 and 26.14 mg GAE·g⁻¹ dry matter for the wild and 14-4 genotype murtilla). Thus, 49.5% ethanol (v/v, ethanol:water) was the optimal solvent concentration for the extraction of this substrate. Values over 50% ethanol (v/v, ethanol:water) showed a reduced extraction efficiency.

Our results show that under the same solvent concentration, the total phenolic content of the 14-4 genotype murtilla fruit extract was significantly higher than that of the wild murtilla fruit extract. This outcome is consistent with the concentration reported by Tsao et al. (2003), who found differences in the phenolic content of a cultivated variety of Canadian strawberries and the wild fresh fruit using the same extraction solvent.

Boeing et al. (2014) evaluated the effect of different solvents and their combinations on the extraction of antioxidant compounds from black mulberries (Morus nigra), blackberries (Rubus ulmifolius) and strawberries (Fragaria x ananassa), and they determined the total phenolic content, anthocyanin content and antioxidant capacity. For all of the berries, the acetone/water solvent mixture (70–30%, v/v) presented the optimal phenolic compound extraction capacity, whereas for black mulberries, the ethanol/water solvent mixture (50/50, v/v) presented the highest antioxidant capacity.

Radojković et al. (2013) evaluated the influence of solvent composition (ethanol/water, 40–80%, v/v), temperature (40–80 °C) and time (20–60 min) on the extraction yield of phenolic compounds, flavonoids, and saccharides and the antioxidant activity of black mulberries (Morus nigra L.). These authors observed that the optimal conditions within the experimental range of the studied variables were as follows: solvent composition of 58.7%, temperature of 58.1 °C and extraction time of 46.9 min.

According to the results of this study, the optimal conditions for obtaining extracts with a higher total phenolic content was the same for the wild and 14-4 genotype fruits under 49.5% ethanol at 30 ºC. Because temperature did not have a significant influence on the amount of total phenolic compounds, a lower temperature was selected to save energy. Therefore, all of the subsequent analyses were performed using extracts produced under optimized conditions (49.5% and 30 ºC) for both samples. Certain components, such as quercetin, epicatechin, and gallic, benzoic and hydrocaffeic acids, were identified by a CG/MS analysis in a previous study (Augusto et al., 2014).